Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Clean the user's REAL Notion database — read it, find stale/incomplete/duplicate entries, and fix them via the connector — not advice on keeping Notion tidy. Use when asked to clean up my Notion database, my tracker is a mess, find the stale and duplicate entries, or tidy my projects DB in Cowork. Reads the database via the Notion connector, audits for staleness / missing required fields / duplicates / status drift, and produces a hygiene-report artifact plus the applied fixes (with a preview-an
.claude/skills/mohitagw15856-notion-db-hygiene/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 43 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 487% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 13% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 75% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -5% | 0% |
Trackers rot: entries stuck "In progress" for months, blank owners, the same project logged twice, statuses that no longer match reality. In Claude Cowork this skill reads the real database, finds the rot, and — after showing you the plan — fixes it in place, so the tracker becomes trustworthy again.
Ask for these if not provided:
Guardrails: never write before the preview is approved; never hard-delete — archive; when merging, preserve the richer record and its relations; respect the user's rules over defaults; if the connector is unauthorised, produce the report and plan without applying, and say so.
A Notion Hygiene Report:
R rows · S stale · I incomplete · D duplicates · X status-drift
| Row | Issue | Proposed fix | |---|---|---|
| Row | Change made | |---|---|
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 3,069 | 10,417 | +239% | 1 | 1 | 0% | 448 | 2,631 | +487% | 0 | 0 | — |
case-02 | fail→pass | 9,955 | 6,667 | -33% | 1 | 1 | 0% | 1,813 | 2,055 | +13% | 0 | 0 | — |
case-03 | pass→pass | 7,617 | 6,450 | -15% | 1 | 1 | 0% | 1,582 | 2,267 | +43% | 0 | 0 | — |
case-04 | pass→pass | 10,866 | 11,224 | +3% | 1 | 1 | 0% | 2,195 | 2,995 | +36% | 0 | 0 | — |
case-05 | pass→pass | 9,014 | 7,671 | -15% | 1 | 1 | 0% | 1,882 | 2,445 | +30% | 0 | 0 | — |
case-06 | fail→pass | 10,769 | 6,223 | -42% | 1 | 1 | 0% | 1,879 | 1,999 | +6% | 0 | 0 | — |
case-07 | fail→pass | 5,718 | 5,015 | -12% | 1 | 1 | 0% | 983 | 1,724 | +75% | 0 | 0 | — |
case-08 | pass→pass | 7,896 | 4,872 | -38% | 1 | 1 | 0% | 1,343 | 1,742 | +30% | 0 | 0 | — |
case-09 | pass→fail | 5,877 | 8,523 | +45% | 1 | 1 | 0% | 946 | 2,191 | +132% | 0 | 0 | — |
case-10 | pass→pass | 8,708 | 5,884 | -32% | 1 | 1 | 0% | 1,493 | 1,791 | +20% | 0 | 0 | — |
case-11 | pass→pass | 5,560 | 3,354 | -40% | 1 | 1 | 0% | 1,006 | 1,458 | +45% | 0 | 0 | — |
case-12 | pass→pass | 4,185 | 2,909 | -30% | 1 | 1 | 0% | 683 | 1,474 | +116% | 0 | 0 | — |
case-13 | fail→fail | 5,464 | 4,193 | -23% | 1 | 1 | 0% | 923 | 1,622 | +76% | 0 | 0 | — |
case-14 | pass→pass | 7,933 | 3,924 | -51% | 1 | 1 | 0% | 1,402 | 1,539 | +10% | 0 | 0 | — |
case-15 | fail→pass | 10,707 | 4,919 | -54% | 1 | 1 | 0% | 1,895 | 1,803 | -5% | 0 | 0 | — |
case-16 | fail→pass | 12,532 | 6,263 | -50% | 1 | 1 | 0% | 1,948 | 1,960 | +1% | 0 | 0 | — |
case-17 | fail→pass | 5,745 | 6,926 | +21% | 1 | 1 | 0% | 932 | 2,117 | +127% | 0 | 0 | — |
case-18 | pass→pass | 10,897 | 6,304 | -42% | 1 | 1 | 0% | 1,850 | 1,991 | +8% | 0 | 0 | — |
case-19 | pass→pass | 5,327 | 4,210 | -21% | 1 | 1 | 0% | 893 | 1,611 | +80% | 0 | 0 | — |
case-20 | pass→pass | 6,068 | 3,776 | -38% | 1 | 1 | 0% | 1,015 | 1,526 | +50% | 0 | 0 | — |
case-21 | pass→pass | 7,773 | 2,777 | -64% | 1 | 1 | 0% | 1,178 | 1,325 | +12% | 0 | 0 | — |
case-22 | pass→pass | 7,066 | 6,680 | -5% | 1 | 1 | 0% | 1,157 | 1,898 | +64% | 0 | 0 | — |
case-23 | pass→pass | 12,109 | 4,488 | -63% | 1 | 1 | 0% | 2,084 | 1,608 | -23% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.